A Short-Cut Method for the Quantification of ... - ACS Publications

A short-cut method for the estimation of nucleation, growth, and dissolution kinetics for one-dimensional population balance equations models is devel...
0 downloads 0 Views 3MB Size
Subscriber access provided by CORNELL UNIVERSITY LIBRARY

Article

A short-cut-method for the quantification of crystallization kinetics - Part 1: Method development Erik Temmel, Holger Eisenschmidt, Heike Lorenz, Kai Sundmacher, and Andreas Seidel-Morgenstern Cryst. Growth Des., Just Accepted Manuscript • DOI: 10.1021/acs.cgd.6b00787 • Publication Date (Web): 17 Oct 2016 Downloaded from http://pubs.acs.org on October 26, 2016

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Crystal Growth & Design is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 32

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Crystal Growth & Design

A short-cut-method for the quantification of crystallization kinetics Part 1: Method development Erik Temmel1, Holger Eisenschmidt2, Heike Lorenz1, Kai Sundmacher1,2, Andreas SeidelMorgenstern1,3* 1 Max Planck Institute for Dynamics of Complex Technical Systems, Sandtorstr. 1, D-39106 Magdeburg, Germany 2 Otto von Guericke University Magdeburg, Chair of Process Systems Engineering, Universitätsplatz 2, D-39106 Magdeburg, Germany 3 Otto von Guericke University Magdeburg, Chair of Chemical Process Engineering, Universitätsplatz 2, D-39106 Magdeburg, Germany Abstract Population balance models are of high interest for the efficient design, control and optimization of crystallization processes. They usually contain mathematical sub-models for the description of the relevant kinetic phenomena, such as growth, dissolution and nucleation of particles. Commonly, component-specific parameters included in these sub-models have to be determined for every substance system of interest or even for the specific experimental set-up. Thus, a short-cut-method is suggested that is based on analyzing the evolution of the crystal size distribution during a few batch crystallization experiments to efficiently parameterize kinetic sub-models required to predict and evaluate the performance of crystallization processes. To illustrate the overall procedure and to evaluate the accuracy of the proposed approach, in silico data with pre-defined kinetics corresponding to hypothetical non-isothermal batch runs are analyzed. It is shown that it is possible to recover pre-specified sub-model parameters with a rather limited amount of input information. Subsequently, less flexible sub-models and measurement errors are considered for comparison, in order to evaluate the loss of predictability. The agreement found between the results of a simulated continuous crystallization process applying both, (a) the initially provided model together with the pre-specified parameters or (b) parameter estimates provided by the short-cut-method demonstrates the practical applicability of the latter.

Corresponding author:

Email:

Andreas Seidel-Morgenstern Max Planck Institute for Dynamics of Complex Technical Systems; Sandtorstraße 1, 39106 Magdeburg, Germany Tel: +49 391 6110401; Fax: +49 391 6110521 [email protected]

1 ACS Paragon Plus Environment

Crystal Growth & Design

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 2 of 32

A short-cut-method for the quantification of crystallization kinetics Part 1: Method development Erik Temmel1, Holger Eisenschmidt2, Heike Lorenz1, Kai Sundmacher1,2, Andreas Seidel-Morgenstern1,3* 1 Max Planck Institute for Dynamics of Complex Technical Systems, Sandtorstr. 1, D39106 Magdeburg, Germany 2 Otto von Guericke University Magdeburg, Chair of Process Systems Engineering, Universitätsplatz 2, D-39106 Magdeburg, Germany 3 Otto von Guericke University Magdeburg, Chair of Chemical Process Engineering, Universitätsplatz 2, D-39106 Magdeburg, Germany * Corresponding author: [email protected] Abstract Population balance models are of high interest for the efficient design, control and optimization of crystallization processes. They usually contain mathematical sub-models for the description of the relevant kinetic phenomena, such as growth, dissolution and nucleation of particles. Commonly, component-specific parameters included in these sub-models have to be determined for every substance system of interest or even for the specific experimental set-up. Thus, a short-cut-method is suggested that is based on analyzing the evolution of the crystal size distribution during a few batch crystallization experiments to efficiently parameterize kinetic sub-models required to predict and evaluate the performance of crystallization processes. To illustrate the overall procedure and to evaluate the accuracy of the proposed approach, in silico data with pre-defined kinetics corresponding to hypothetical non-isothermal batch runs are analyzed. It is shown that it is possible to recover pre-specified sub-model parameters with a rather limited amount of input information. Subsequently, less flexible sub-models and measurement errors are considered for comparison, in order to evaluate the loss of predictability. The agreement found between the results of a simulated continuous crystallization process applying both, (a) the initially provided model together with the pre-specified parameters or (b) parameter estimates provided by the short-cut-method demonstrates the practical applicability of the latter. Keywords: Crystallization processes, Particle size distributions, Kinetics, Nucleation, Growth, Dissolution 1.

Introduction

Many industrially important substances are produced with a tailored product quality via crystallization. Ordinary sugar, for example, is desired to have a narrow crystal size distribution (CSD) with a mean value of 700-800 µm, almost without fine material and a 2 ACS Paragon Plus Environment

Page 3 of 32

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Crystal Growth & Design

purity of 99.5% [1]. This is feasible exploiting one single seeded batch process based on the experience accumulated over many decades. For current industrial applications, faster approaches are needed to design crystallization processes. Increasingly theoretical concepts based on population balance equations (PBE) are utilized for this purpose. However, it is difficult to quantify the corresponding rates, e.g. for growth, nucleation and dissolution, which are required for this mathematical framework [2]. Several publications describe a rather simple method, based on the observation of the evolution of the solid and liquid phase during well planned cooling batch crystallization in order to quantify the necessary crystallization kinetics. Obviously, crystal populations move to larger particle sizes due to growth (G) and to smaller sizes due to dissolution (D) (Fig. 1a and 1b, [3, 4]). It was theoretically shown, e.g. in [5], that it is possible to determine the growth rate by evaluating one specific characteristic of the CSD during crystallization. Hence, it is straightforward to exploit seeded batch crystallization experiments for this purpose if size-independent growth [6] is assumed even though growth rate dispersion occurs. The crystals, which are initially provided with an appropriate size and shape, can be observed during the crystallization process utilizing suitable online or offline measurement techniques, as e.g. sieve analysis or online microscopy. Therefore, the seed particles should be large enough to distinguish them from eventually occurring nuclei and the substance-specific crystal shape should be developed completely to reduce their agglomeration tendency. Based on their specific growth or dissolution rates, the crystals move during the process with a certain velocity through the space of the characteristic coordinates (e.g. length or width of the particles). Thus, it is instructive and valuable to quantify the crystal size evolutions since they contain information regarding the underlying growth or dissolution kinetics. This information can be utilized directly (without the need of solving full population balance equations, PBE) together with observed supersaturation and temperature in the crystallizer to estimate free parameters in postulated kinetic sub-models for growth or dissolution. A similar simple procedure can be applied to analyze observed appearance (birth) rates of new crystals (B0). Observed positive temporal changes in the total number of crystals (N(t), Fig. 1c), are directly connected to the nucleation kinetics provided that breakage and agglomeration events are negligible.

Fig. 1: a) & b): Movement of a seed-peak due to crystal growth (G) and dissolution (D). c): Change of the particle number (N(t)) of the CSD due to nucleation (B0) [7]. Solid lines: Initial state; Dashed lines: Evolution of the state.

3 ACS Paragon Plus Environment

Crystal Growth & Design

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 4 of 32

Altogether, the described evaluation procedure based on classical experiments leads to an efficient and fast quantification scheme. Essential information is extracted from measured CSDs and applied directly for the parameterization of different postulated kinetic models. The just described simple (“PBE-free“) procedure will be designated in the following as “Short-cut-method”. In recent years, several authors performed investigations around this main idea. One work describes the experimental set-up and data processing to calculate nucleation and growth rates during seeded cooling crystallizations at fixed supersaturation [8]. In that work, laser diffraction was applied to investigate the solid phase that can lead to errors when the crystal shape differs significantly from perfect spheres since reflections of facetted particles and the orientation of crystals influence the result significantly [9, 10]. Furthermore, the mean length of the whole crystal size distribution (CSD) was taken as an indicator for the growth of the particles [8]. This eventually obscures the real growth rate, in particular if nucleation occurs. Yokota et al. [11] analyzed four precisely determined CSDs of one non-isothermal batch crystallization and identified successfully the three parameters (kg,0, EA,g, g) of an assumed temperature-dependent power law for the crystal growth kinetics (eq. 1) G = k g ,0 exp(

− EA, g RT

(eq. 1)

)( S − 1) g

It is however rather difficult to accurately measure CSDs during an experiment. Consequently, four determined crystal size distributions may not be sufficient. In this paper, the potential of the short-cut-method is investigated more systematically utilizing predicted transients of batch experiments. The approach of [11] is extended in order to identify besides growth, also nucleation and dissolution kinetics and to evaluate the impact of the number, type and precision of input data on the quality of the estimated kinetic parameters. It is shown, that only a few experiments are necessary to estimate all basic crystallization kinetics for a large range of supersaturation and temperature. Furthermore, the numerical effort is reduced since no full PBE model is applied and only the parameters of one postulated kinetic sub-model is quantified at once. Finally the applicability of the short-cut method is also tested by comparing design scenarios for a continuous crystallization process using various parameter constellations. This article is organized as follows: In the following section, the model framework that was applied for the generation of synthetic (“perfect”) crystallization data is explained together with the properties of the assumed substance system. Subsequently, a synthetic reference data set of crystallization experiments is created. In the third section, the short-cut-method is explained in detail and the analysis of the synthetic crystallization experiments is presented. In the fourth section, the parameter optimization is shown for different amounts of data with and without noisy measurement signals. In the last section, a continuous crystallization process is designed with the estimated kinetic rates and with the initially provided values for comparison. Finally, all results are evaluated in a concluding section.

4 ACS Paragon Plus Environment

Page 5 of 32

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Crystal Growth & Design

2.

Generation of in silico process data

For the evaluation of the short-cut-method, in silico data of batch crystallization experiments are generated and subsequently analyzed. A one-dimensional PBE system is used to create the required information. This ideal case assumes (at first) perfect concentration and temperature signals as well as a highly accurate measurement of the size distribution during the experiment. The model framework and all necessary process and kinetic parameters are explained in the following.

2.1. Model framework For the crystallization case, the general population balance reduces to equation 2 if breakage and agglomeration are neglected, size independent growth and ideal mixing without volume contraction due to an occurring solid phase are assumed [12]. ∂f (t , L) ∂f (t , L) = −G ( S (t ), T (t )) ∂t ∂L

for S ≥ 1

(eq. 2)

Here, L is the internal size coordinate and G represents the growth rate. The corresponding driving force for crystal growth in equation 2, is the supersaturation, which is defined for ideal solutions [13] as the ratio between the actual concentration, c, and the saturation concentration, c*(T), at a certain saturation temperature.

S=

c c (T )

(eq. 3)

*

For the simple batch case, the corresponding mass balance consists of the accumulation in the liquid phase and the transport of substance to the solid phase due to crystallization.

3k ρ G( S (t ), T (t )) L max 2 dc(t ) = − v solid ∫ L f (t, L)dL dt Vsolvent ρsolvent L min

(eq. 4)

It should be noted that a constant solvent mass is assumed in equation 4. The population balance was subsequently discretized applying a finite volume method together with the upwind scheme [14], since an analytical solution is not available for this PDE system. For the resulting ODE system, three initial conditions (IC) are required. It is assumed that the seed crystals can be characterized by a perfect Gaussian distribution (eq. 5) with a certain mean value, LSeed , and standard deviation, σSeed. Even though the same seed mass is used throughout the simulations, the initial distribution is scaled to a certain mass, mSeed, with equation 6 since the Gaussian distribution is normalized.

5 ACS Paragon Plus Environment

Crystal Growth & Design

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

1

f 0 , Seed ( L ) =

σ Seed

 1 L−L   Seed exp  −     2 σ 2π Seed    

f Seed ( L) = f 0, Seed ( L) ⋅

Page 6 of 32

2

mseed

(eq. 6)

L max

kv ρ solid



(eq. 5)

3

L f 0, Seed ( L)dL

L min

Furthermore, it is assumed that the liquid phase is saturated at a certain temperature, T0, with a concentration c0 (eq. 8). The occurrence of new crystals due to nucleation is introduced as an initial condition at the crystal size L = 0 (eq. 9). Subsequently, all conditions are defined for the crystallization case as follows: IC for S > 1:

f (t = 0, L) = f Seed ( L)

(eq. 7)

c(t = 0) = c 0 = c* (T0 )

(eq. 8)

f (t , L = 0) =

B0 G

(eq. 9)

Similar to [11] (eq. 1), temperature and supersaturation dependent power-law approaches with three parameters (eq. 10) are applied to quantify the rates of the specific phenomena considered in equations 2, 4 and 9. K = p1 exp(

− p2 )( S − 1) p3 RT

for K = G, D, B0

(eq. 10)

Here, p1 is the pre-exponential factor, k0, p2 is usually referred to as the activation energy, EA, and p3 is the power law exponent. The in silico batch crystallization experiments were predicted as a combination of crystallization and dissolution to have information about all three kinetic phenomena considered in this study within one experiment. At first, after seeding the crystallizer is cooled down linearly until a certain temperature difference is achieved. Hence, the driving force will raise and nucleation and growth of present crystals will take place. Subsequently, the reactor is heated again with a linear temperature ramp back to the starting temperature. The dissolution rate, D, replaces the growth rate, G, in equations 2, 4 and 9 in case of undersaturation (S